Table 4.

Results of multiple variable linear regression analyses using all clinical and pathological variables for each research cohort separately. The research cohorts were analyzed separately first. Note that the linear regression found large differences between the cohorts in the y‐intercept, and in each case the isocortical NFTs was the largest parameter estimate. Other regressors differed, which may indicate biological differences in the parameters that affect relatively low‐CILA and high‐CILA individuals. Regressors with positive parameter estimates were not listed on this table. There are no such regressors in the “initially normal” group but ApoE and presence of diabetes were positive parameter estimates in the “dementia clinic” group. Abbreviations: ApoE = apolipoprotein E; CILA = cognitive impairment linked to aging; HS = hippocampal sclerosis; LB = Lewy body; NFT = neurofibrillary tangle; TIA = transient ischemic attack.

Parameter	Parameter estimate	Standard error	Pr > |t|
“Initially normal” group n = 212
Intercept	29.49	0.31	<0.0001
NFTs, isocortex	−24.93	4.30	<0.0001
NFTs, mesial temporal	−16.10	2.09	<0.0001
HS—unilateral	−12.53	1.48	<0.0001
HS—bilateral	−5.50	1.65	0.001
LBs, mesial temporal	−5.61	1.56	0.0004
LBs, amygdala only	−3.73	1.51	0.0146
History of TIA	−1.55	0.72	0.0334
“Dementia clinic” group, n = 122
Intercept	10.47	2.05	<0.0001
NFTs, isocortex	−16.42	2.42	<0.0001
LBs, neocortex	−5.07	2.10	0.0177
Antipsychotics	−4.61	2.01	0.0237
Seizures	−4.31	1.91	0.0257
Cancer	−2.84	1.40	0.0447